Dunning Management: The Complete Playbook for Recovering Failed Payments

Most subscription brands treat dunning management as a billing setting. Somebody turned it on during implementation, picked whatever retry schedule the platform suggested, wrote three emails in an afternoon, and nobody has looked at it since. Meanwhile it is quietly deciding what percentage of your MRR survives each month.

I run a done-for-you phone recovery service. We come in after the automated layer has finished its run, and the first thing we ask for is the failed-charge file. What is in that file tells you more about a brand’s real retention than any survey. A brand with a 7% monthly failed-charge rate and a mediocre dunning process is bleeding revenue from subscribers who never decided to leave.

That is the whole point of this piece. Dunning is not a billing chore. It is a revenue function with an owner, a target, and a weekly number.

What dunning management actually is

Dunning is everything that happens between a renewal charge failing and that subscriber either paying or being written off.

It has four moving parts:

Retries. Re-presenting the same card on a schedule, hoping the condition that caused the decline has changed.
Account updater. Pulling refreshed card credentials from the card networks when the subscriber got a new card.
Messaging. Email and SMS asking the subscriber to update their payment method, with a link that makes it easy.
Escalation and exit. What you do at the end of the sequence, and when you finally cancel the subscription.

Card issuers will tell you nothing useful about a given decline beyond a code. Your job is to read that code correctly and pick the right one of those four levers. Almost every bad dunning process I have seen fails because it applies the same lever to every failure.

The other thing to be clear about up front: failed-payment churn is not the same as someone choosing to cancel. It is a different problem with a different fix, and it deserves its own line in your churn reporting. I go into the distinction at length in our piece on involuntary churn, and the broader concept of churn is covered in what customer churn is and how to measure it. This post is about the tactics.

Why payments fail, and why the reason dictates the fix

Every processor returns a decline code. Stripe, Adyen, Braintree, all of them. The codes are not standardized across the industry in any clean way, but they cluster into a handful of categories, and those categories are what your dunning logic should branch on.

Soft declines

Soft declines are temporary conditions. The card is valid, the account is valid, the charge just did not go through right now.

Insufficient funds. The most common soft decline in consumer subscriptions. It is a timing problem, not a willingness problem. Retrying this one on the right day is genuinely effective, and it is the single best argument for retry timing that tracks payday cycles rather than a flat 3-day interval.
Temporary holds and velocity limits. The issuer throttled the card because of activity patterns. Usually clears on its own.
Processor or network errors. Something timed out. Retry.

For soft declines, retry is the right lever and messaging is the wrong first move. Emailing a subscriber “your payment failed” when the charge will succeed on its own in four days generates support tickets and cancellations you did not need to create.

Hard declines

Hard declines mean the card, as presented, will never work again.

Card expired. The subscriber has a new card in their wallet. Your system has the old number. No amount of retrying fixes this.
Stolen or lost card, account closed. Same story.
Do not honor. This is the ugly one because it is a catch-all. The issuer declined and did not say why. Sometimes a retry works, often it does not.
Issuer fraud rules. The issuer flagged the merchant, the amount, or the pattern. Retrying the identical charge repeatedly makes this worse, not better.

For hard declines, retries are wasted attempts and messaging is the only lever that matters. You need the subscriber to physically give you a new payment method.

The friction cases

Two more failure modes that are not really declines but show up in the same bucket:

3DS and SCA friction. For subscribers in Europe and the UK, strong customer authentication rules mean some recurring charges get kicked back for authentication. Recurring transactions have exemptions available, but if your setup is not using them correctly you generate failures that look like declines and are actually configuration.
Network account updater gaps. Visa and Mastercard both run services that push updated card credentials to merchants when a card is reissued. They are excellent and they are not complete. Not every issuer participates fully, not every card type is covered, and updates do not arrive instantly. A meaningful share of expired-card failures are ones the updater should have caught and did not.

The practical takeaway: route your dunning logic by decline category. Soft declines get retries first, hard declines get messaging first, authentication failures get a configuration fix rather than either.

Payment retry strategy that does not annoy issuers

Retry strategy is where most of the cheap recovery lives, and where most brands under-think it.

Retry timing beats retry count. More attempts on a card that will never work adds nothing. Fewer, better-timed attempts on a card that might work adds a lot. If your subscriber base is consumer and US-based, the first and fifteenth of the month plus Fridays are not a coincidence, they are when accounts get funded. Aligning insufficient-funds retries to that reality is the highest-leverage change most brands can make to their retry schedule.

Cap the attempts. Card networks and issuers do monitor retry behavior, and excessive re-presentment of declined transactions can carry penalties depending on the network and the scheme rules in force. Beyond the compliance question, hammering a card that returned a hard decline just trains the issuer’s fraud model to distrust you. Most platforms default to somewhere between three and five attempts over two to three weeks. If you want more attempts than that, get a real reason first.

Vary the amount only if it makes sense for your model. Some brands retry at a partial amount for insufficient-funds declines, particularly on higher-ticket subscriptions. It can work. It also complicates your revenue reporting and your customer service conversations. Test it, do not assume it.

Run account updater before you run retries on expired cards. If a card is expired and the network has a refreshed credential, you want that credential in place before you attempt anything. Most modern billing platforms support this. Check that yours is actually enabled, because I have seen plenty of accounts where it was available and switched off.

Pre-dunning is underrated. The cheapest failed payment is the one that never happens. Cards carry an expiry date you already have. Flag the ones expiring inside the next 30 days and message those subscribers before the charge. Same for a renewal notice a few days ahead on higher-ticket plans. You are not asking for anything except a quick card check, and you convert a would-be involuntary cancellation into a non-event.

The dunning email and SMS sequence, done properly

Once you know a charge needs human action, the messaging sequence is what carries the load. A few rules I would defend.

One-click update, always. The single biggest determinant of whether a dunning email works is how many steps stand between the click and the new card being saved. A link that drops the subscriber into a logged-in, pre-authenticated update page converts. A link that drops them on a login screen where they have to remember a password loses most of them right there. If you fix one thing in your dunning emails this quarter, fix this.

Tone is service, not collections. The subscriber’s card stopped working. That is all that happened. They are not delinquent, they did not do anything wrong, and the moment your copy implies otherwise you have turned a billing hiccup into a reason to reconsider the subscription. “Your card was declined, update it or your subscription will be cancelled” reads as a threat. “Looks like your card on file expired, here is a link to update it so your next box ships on time” reads as help.

Lead with the consequence they care about, not the one you care about. They do not care that your charge failed. They care that their order is not shipping, their access is paused, or their delivery date is slipping. Say that.

A workable cadence. Roughly:

Day 0, on failure: short, friendly, one-click link. No urgency language. For soft declines, consider suppressing this one entirely and letting the retry run first.
Day 3: second email, slightly more specific about what happens next. Add SMS here if you have consent, because SMS out-performs email on this job for most consumer subscription brands and the message is short by nature.
Day 7: state the deadline plainly. Give the date the subscription pauses.
Day 10 to 14: final notice, plus a genuine alternative. If a subscriber cannot pay right now, offering a pause or a cheaper plan keeps the relationship alive instead of ending it.
End of sequence: cancel or park the subscription per your policy.

Send from a real address and make replying work. Dunning email that comes from noreply@ and bounces replies is throwing away the most engaged subscribers in the sequence, the ones who wrote back.

Check deliverability separately from the copy. Transactional billing email usually enjoys good inbox placement, but not if it is going out on a shared marketing IP with your promotional sends. Worth a look if your open rates on dunning are suspiciously low.

What platform dunning does and does not cover

I want to be fair here, because the strawman version of this argument is dishonest and operators can smell it.

Recharge has a full dunning engine, and Recharge Concierge adds a managed outreach layer on top of it. Skio and Stay AI both ship dunning and retry logic built for Shopify subscription brands, with their own approaches to retry timing and messaging. Stripe runs Smart Retries, which uses network-level data to pick retry timing rather than a fixed schedule, plus configurable dunning email.

These are good products built by people who understand payments. In our own engagements, the brand’s existing flows always run first and get their complete run. We do not replace them, and I would not recommend any brand replace a working dunning engine with a phone call.

But here is what every one of these systems has in common: they work through digital channels, on a fixed sequence, and when the sequence ends they stop. What they cannot do is get a response out of someone who is not opening email. And at the end of a typical dunning sequence, that describes most of who is left. The easy recoveries already happened. The remaining pool is disproportionately expired and reissued cards belonging to subscribers who intended to keep the subscription and simply have not gotten around to it.

That is not a failure of the software. Software has a ceiling on this job and the ceiling is real.

The gap at the end, and the human call that closes it

This is Play 02 in what we run for ecommerce brands, and the trigger is specific: a renewal charge has failed and the platform’s dunning has finished without recovering it.

At that moment the subscriber is sitting in a queue waiting to be cancelled. A trained agent calls them. The conversation is not collections and it is not a save call. It is roughly thirty seconds of substance: your card on file stopped working, we did not want your order to lapse, can we update it now. Most of the people who pick up say yes, because they were never trying to leave.

What counts as recovered is the first charge after the payment method is updated. Not an intent, not a promise, an actual settled charge. We price on performance, so if the charge does not go through there is nothing to bill for. You can see the full mechanics on the payment recovery page and how it sits alongside the other plays on the ecommerce page.

When a human call is worth it, and when it is not

I will say the unhelpful-to-me part out loud, because there is no version of this where calling everyone makes sense.

It does not pencil when:

The ticket is small. A $19 monthly subscription where the recovered charge is a single $19 has to survive a lot of failed dials before it pays for the connected one. Low-ticket brands are better served by tightening retry timing and the one-click update flow.
LTV is short. If the average subscriber churns voluntarily after two more cycles anyway, the recovered charge is worth two cycles, not twelve. Recovery economics depend on what happens after the save, not just the save.
The failed-charge pool is tiny. A few dozen failures a month does not support a calling operation of any kind, ours or in-house. Fix the automated layer first.
Your automated layer is genuinely broken. If your dunning emails go to a login wall, or account updater is switched off, calling people is an expensive way to paper over a cheap fix. Do the cheap fix.

It does pencil when: the ticket is meaningful, the subscription renews for a long time, the failed-charge pool is large enough to be a real queue every week, and the automated layer is already good. Which is exactly why we insist the brand’s own flows run first. We want the pool we work to be the genuinely hard residue, not the easy recoveries someone’s email would have gotten anyway.

If you want the general argument about where automation stops and people start, human versus AI reactivation covers it across other channels.

Measuring dunning management

You cannot run this as a revenue function without a scoreboard. Four numbers.

1. Recovery rate. Of the charges that failed in a given month, what percentage eventually settled? Segment it by decline category, because a blended number hides everything useful. Soft-decline recovery and expired-card recovery behave completely differently and if you only track the blend you will never know which half of your process is weak.

2. Days to recover. How long from first failure to settled charge? This is your cash-flow number and it is also an early warning. If days-to-recover is creeping up, something in the early sequence has stopped working.

3. Failed-payment churn as a share of total churn. Split your cancellations into “the subscriber chose this” and “the billing system did this.” Most brands are surprised by the split the first time they run it. If billing-driven cancellations are a large share, dunning is your cheapest retention project, full stop. More on the overall picture in subscription churn.

4. What counts as recovered, defined once and written down. This is where most reporting arguments come from. Our definition, for the work we do: the first charge that settles after the payment method is updated, inside the agreed attribution window. Exclusions stated up front: refunds, chargebacks, and charges that would have recovered on their own without any outreach. Whatever definition you pick, pick one and hold everyone, including any vendor, to the same one.

Also worth tracking as a health check: retry attempts per recovered charge. If that number is climbing, you are buying recoveries with attempts rather than with better targeting.

A note on compliance

Short, and not legal advice. Talk to your own counsel.

Calling subscribers about a failed payment in the US sits under TCPA and the federal Do Not Call framework. The practical points: the existing business relationship with a paying subscriber matters and is generally the basis for this kind of contact, calls need to stay inside permitted calling hours in the subscriber’s local time zone, DNC status has to be respected and checked rather than assumed, and opt-outs get honored immediately and permanently. Same for SMS, where consent scope is its own conversation and a marketing consent is not a billing consent.

None of this is exotic, but it is the part where doing it casually creates real exposure. Whoever runs your calling, in-house or outsourced, should be able to tell you exactly how they handle each of the above without checking.

An illustrative model

These numbers are made up to show the shape of the math. They are not benchmarks, they are not our results, and they are not a forecast for your brand. Put your own numbers in.

A hypothetical subscription brand:

Input	Illustrative value
Active subscribers	10,000
Average order value per renewal	$60
Monthly renewal revenue	$600,000
Renewal charges that fail in a month	7% (700 charges)
Revenue at risk from failures	$42,000
Recovered by retries, updater and dunning messaging	60% of the at-risk pool
Revenue recovered by the automated layer	$25,200
Remaining pool after dunning finishes	$16,800 (280 charges)

Now model the human layer on that leftover pool at three different assumed recovery rates, because the honest thing to do with a number you cannot know in advance is to show a range:

Assumed recovery on the leftover pool	Recovered per month	Annualized
15%	$2,520	$30,240
25%	$4,200	$50,400
35%	$5,880	$70,560

And the point that most models miss: the recovered charge is not the whole value. A subscriber saved in month one keeps renewing. At a $60 AOV, a saved subscriber who goes on to renew six more times is worth $360, not $60. That multiple is what decides whether calling pencils, and it is why short-LTV brands should not bother and long-LTV brands should.

If you want to run this against your actual numbers rather than mine, the ROI calculator does the arithmetic. And if dunning is one part of a broader retention problem, retention strategies for service businesses and our guide to abandoned cart recovery cover the adjacent pieces.

Where to start this week

If you want a sequence rather than a list:

Pull 90 days of failed charges and group them by decline category. You will probably learn something in the first ten minutes.
Check whether account updater is actually enabled. Not available, enabled.
Click your own dunning email from a logged-out phone browser and count the steps to a saved card. Fix whatever you find.
Re-time your insufficient-funds retries around funding cycles instead of a flat interval.
Turn on a pre-dunning notice for cards expiring in the next 30 days.
Only then look at what is left at the end of the sequence and decide whether it is big enough to call.

Steps one through five are free and most brands have not done them. Step six is where we come in, and only if the first five are already done.

FAQ

What is dunning management?

Dunning management is the process of recovering failed recurring payments. It covers retrying declined charges on a schedule, pulling updated card credentials from the card networks, sending dunning emails and SMS asking the subscriber to update their payment method, and deciding when to pause or cancel a subscription that never recovers. Done well it is a revenue function with its own owner and metrics, not a billing setting.

How many times should you retry a failed payment?

Most billing platforms default to three to five attempts spread over two to three weeks, and that is a reasonable starting point. Timing matters more than count. Retries on insufficient-funds declines do better when aligned to when accounts are typically funded, and retries on hard declines like expired or closed cards are wasted attempts regardless of timing, since the card will never work again.

What is the difference between a hard decline and a soft decline?

A soft decline is temporary. Insufficient funds, a velocity limit, a network timeout. The card is valid and a later retry may succeed. A hard decline means the card as presented will not work again: expired, closed, reported lost or stolen. Soft declines are solved by retrying. Hard declines are only solved by getting a new payment method from the subscriber.

How many dunning emails should a sequence have?

Three to five over roughly two weeks works for most consumer subscription brands, starting friendly and becoming specific about the pause date as the sequence progresses. Adding SMS where you have proper consent generally lifts response. The number of emails matters far less than whether the update link takes the subscriber straight into a pre-authenticated card-update page in one click.

Do you replace our existing dunning platform?

No. Recharge including Concierge, Skio, Stay AI and Stripe retries all get their full run first. We work the charges that are still unrecovered when that sequence ends, by phone, and what counts as recovered is the first charge that settles after the payment method is updated. We price on performance, so if nothing settles there is nothing to bill.

Dunning management is one of the few retention projects where the revenue is already yours and you are simply failing to collect it. No new acquisition, no discounting, no new product. Just a process that either works or quietly does not.

If you want to know how much is sitting at the end of your sequence, book a strategy call and we will look at your failed-charge file together.